Celebrity Name Conjunction Fallacies
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/bowen-upenn/llm_token_bias
下载链接
链接失效反馈官方服务：
资源简介：
该数据集是一个综合性的合成数据集，旨在评估名人名字在与连珠错觉问题结合时对大型语言模型性能的影响。该数据集突显了名人名字如何可能导致大型语言模型进行错误的推理。该数据集的规模为100个样本，其任务是评估大型语言模型的推理能力。
This is a comprehensive synthetic dataset designed to evaluate the impact of celebrity names when combined with conjunction illusion problems on the performance of large language models (LLMs). This dataset highlights how celebrity names can lead LLMs to engage in erroneous reasoning. Comprising 100 samples, this dataset is utilized to assess the reasoning capabilities of large language models.
提供机构：
Authors of the paper
